ByteDance, the company behind the popular social media platform TikTok, is taking a significant step in artificial intelligence technologies. The company currently relies heavily on NVIDIA’s AI chips but aims to start mass production of its own AI chips by 2026. These chips will be used for two main purposes: artificial intelligence training and artificial intelligence inference. AI training will involve teaching machine learning models containing large data sets, while AI inference will include making real-time predictions like personalized content recommendations on TikTok.
ByteDance’s Plan to Develop Its Own AI Chips
This year, ByteDance spent approximately $2 billion ordering 200,000 H20 GPUs from NVIDIA, but most of these GPUs have not yet been delivered. Facing high costs and limited supply, ByteDance aims to reduce this dependency by developing its own AI chips. The company will collaborate with experienced AI chip designer Broadcom and will use TSMC’s technologies for chip production.
ByteDance’s own GPUs may not be as powerful as NVIDIA’s top-tier models, but the company hopes this move will both save costs and provide more control over the technology. However, they will face significant challenges, such as developing a new software platform to replace NVIDIA’s CUDA software platform.
Expected Changes in China with ByteDance’s AI Chip Production
If ByteDance is successful, fundamental changes could occur in China’s AI hardware market, potentially reducing the country’s dependency on foreign suppliers. However, competing with NVIDIA, a leader in this field, will take time and considerable effort.